1. Field of the Invention
The present invention relates to a system, apparatus, method, and computer program for wireless communication to implement a wideband radio transmission between a plurality of wireless nodes or terminals, as typically performed in a wireless Local Area Network (LAN). In particular, this invention relates to such a system, apparatus, method, and computer program for wireless communication that enhance communication capacity by carrying out Multi Input Multi Output (MIMO) communication using multiple logical channels formed between a pair of a transmitter with multiple antennas and a receiver with multiple antennas by exploiting spatial multiplexing.
More specifically, this invention relates to such a system, apparatus, method, and computer program for wireless communication that carry out closed-loop MIMO transmission, using singular value decomposition (SVD) of a channel matrix with elements that are channels for pairs of transmit antennas and receive antennas, and, in particular, to the system, apparatus, method, and computer program for wireless communication that implement SVD-MIMO communication with a less number of high-load calculation required for SVD processing for a channel matrix.
2. Description of Related Art
Sharing information resources and equipment resources can efficiently be implemented by computer networking typified by LAN. Now, wireless LANs draw attention as systems that free users from cable wiring in traditional wired LANs. Because the wireless LANs can dispense with most of wiring cables in working spaces such as offices, communication terminals such as personal computers (PCs) can be moved more easily.
Recently, because of enhanced speed and reduced price of wireless LANs, the demand for wireless LANs is increasing significantly. Particularly, people think that they will adopt a personal area network (PAN), that is, build a small-scale wireless network with a plurality of electronic devices that are commonly used in living environments for information communication. According to Japanese Radio Regulations, different wireless communication systems and devices can be used in certain frequency bands, e.g. 2.4 GHz and 5 GHz bands, which are permitted for use without a license from the supervisory authority.
Typical standards for wireless networking include IEEE (the Institute of Electrical and Electronics Engineers) 802.11 (e.g., see non-patent document 1), HiperLAN/2 (e.g., see non-patent document 2 or non-patent document 3), IEEE 302.15.3, Bluetooth communication, etc. As for the IEEE 802.11 standard, there are also its enhanced versions IEEE 802.11a (e.g., see non-patent document 4), 802.11b, and 802.11g for different wireless communication systems and frequency bands employed.
The IEEE 802.11a standard supports a modulation scheme achieving a communication speed of up to 54 Mbps. However, there is a need for additional standards that can realize a higher bit rate of communication speed. For instance, IEEE 802.11n aims to develop a wireless LAN technology that allows for an effective throughput higher than 100 Mbps and to establish next-generation wireless LAN standards.
The MIMO communication attracts attention as one technology for realizing higher-speed wireless communication. The MIMO technology achieves enhancement in transmission capacity and communication speed, base on a MIMO system where both transmitter and receiver have multiple antenna elements, thus creating spatially multiplexed transmission channels (hereinafter referred to as “MIMO channels”). The MIMO communication makes an efficient use of a frequency band, because it exploits spatial multiplexing.
The MIMO communication scheme is such that separate transmit data streams are allocated to sub-carriers on the multiple antennas at the transmitter, the sub-carriers are transmitted over multiple virtual MIMO channels, and at the receiver, the sub-carriers received by its multiple antennas are processed and decoded into receive data. This communication scheme exploits channel characteristics, unlike simple adaptive antenna arrays at the transmitter and the receiver.
The MIMO communication system is conceptually depicted in FIG. 4. As shown here, both the transmitter and the receiver are equipped with multiple antennas. At the transmitter, multiple transmit data streams are space-time coded, multiplexed, and allocated to the sub-carriers on M antennas, and the sub-carriers transmitted over multiple MIMO channels. At the receiver, the sub-carriers received by N antennas via the channels are space-time decoded into receive data. This channel model consists of a radio environment (transfer function) around the transmitter, a channel space structure (transfer function) and a radio environment around the receiver (transfer function) Although crosstalk takes place when a transmit signal is multiplexed into sub-carriers and the sub-carriers are transmitted from the transmit antennas, the received multiplexed sub-carriers can be separated into respective proper data streams without crosstalk through signal processing at the receiver.
While various schemes of MIMO transmission configuration have been proposed, it is a significant problem in implementation how to communicate channel information between the transmitter and the receiver, according to the configuration of the antennas.
To communicate channel information, a method of transmitting known information (preamble information) in only one way from the transmitter to the receiver is easy. In this case, the transmitter and the receiver perform spatial multiplexing transmission independently from each other; this is called an open-loop MIMO transmission scheme. An evolved style of this method is a closed-loop MIMO transmission scheme in which ideal, spatially orthogonal channels are created by feedback of preamble information from the receiver to the transmitter as well.
An example of the open-loop MIMO transmission scheme is a Vertical Bell Laboratories Layered Space Time (V-BLAST) scheme (e.g., see patent document 1). The transmitter simply multiplexes a signal into sub-carriers on each transmit antenna and transmits the sub-carriers without assigning a matrix of antenna weighting factors to the sub-carriers. In other words, a procedure of feedback for acquiring the matrix of antenna weighting factors is dispensed with. The transmitter inserts a training signal that is used for channel estimation into the data stream on each antenna, e.g., in a time division manner, before transmitting multiplexed sub-carriers. On the other hand, at the receiver, a channel estimation section performs channel estimation, using the training signal, and calculates a channel information matrix H for every antenna pair. By way of combining zero-forcing and cancellation neatly, the receive signal SN ratio is enhanced taking advantage of spatial degrees of freedom offered by the antennas resulting from the cancellation and decoding probability is increased.
As an ideal form of the closed-loop MIMO transmission, an SVD-MMO scheme using the SVD of a propagation path function is known (e.g., see non-patent document 5).
An SVD-MIMO transmission system is conceptually depicted in FIG. 5. In the SVD-MIMO transmission, UDVH is obtained by singular value decomposition of a numeric matrix with elements of channel information per antenna pair, namely, a channel information matrix H. V as a matrix of antenna weighting factors at the transmitter is assigned to sub-carriers on transmit antennas and UH as a matrix of antenna weighting factors at the receiver is assigned to the sub-carriers received by receive antennas. Consequently, the MIMO channels are represented as a diagonal matrix D with diagonal elements that are square roots of eigenvalues λi per channel, and the multiplexed sub-carriers of a signal can be transmitted without suffering from crosstalk at all. In this case, on both transmit and receive sides, logically independent multiple channels formed by space division, or exactly, spatially orthogonal multiplexing can be realized.
By the SVD-MIMO transmission scheme, it is possible to achieve maximum communication capacity in theory; for instance, if the transmitter and the receiver each have two antennas, the transmission capacity will be doubled at maximum.
The mechanism of the SVD-MIMO transmission scheme is now be discussed in detail. If the transmitter has M antennas, transmit signal x is represented as a set of M×1 vectors; if the receiver has N antennas, receive signal y is represented as a set of M×1 vectors. In this case, channel characteristics are represented as a numeric matrix of N×M, namely, channel matrix H. An element hij of the channel matrix H corresponds to a transfer function from the j-th transmit antenna to the i-th receive antenna. Receive signal vector y is obtained by multiplying the channel information matrix by the transmit signal vector and adding noise vector n to the product, as expressed in the following equation (1).y=Hx+n  (1)
The above-mentioned singular value decomposition of the channel information matrix H is expressed by the following equation (2).H=UDVH  (2)
Here, the matrix V of antenna weighting factors at the transmitter and antenna weight matrix U at the receiver are unitary matrices fulfilling the following equations (3) and (4), respectively.UHU=I  (3)VHV=I  (4)
Specifically, a set of normalized eigenvectors of HHH corresponds to the antenna weight matrix UH at the receiver and a set of normalized eigenvectors of HHH corresponds to the antenna weight matrix V at the transmitter. D is a diagonal matrix with diagonal elements that are square roots of eigenvalues λi of HHH or HHH. The matrix size is determined by the number of transmit antennas M or the number of receive antennas N, which is smaller; that is, a square matrix of size of min (M, N) is obtained and the diagonal matrix is obtained from the square matrix.
                    D        =                  [                                                                                          λ                    1                                                                              …                                                                                                                          0                                                                    ⋮                                                                                  λ                    2                                                                                                                                                                                                                                                                                                                                                                                                                              ⋰                                                                                                                                                  0                                                                                                                                                                                                                                              λ                                          min                      ⁡                                              (                                                  M                          ,                          N                                                )                                                                                                                          ]                                    (        5        )            
Although singular value decomposition for real numbers has been discussed above, care should be taken for singular value decomposition extension up to imaginary numbers. Although U and V are matrices consisting of eigenvectors, eigenvectors with different phases, which are not singular, exist in countless numbers, even if the eigenvectors are manipulated so that a norm of 1 is obtained, in short, they are normalized. In some phasic relationship between U and V, the above equation (2) is dissatisfied, because the phases of U and V are angled differently, though both the U and V are valid. For complete phase matching, V is obtained normally as set of eigenvectors of HHH. However, U is obtained by multiplying the both sides of the above equation (2) by V, as expressed in the following equation.HV=UDVHV=UDI=UD U=HVD−1  (6)
The transmitter transmits sub-carriers weighted by the matrix V of transmit antenna weighting factors and the receiver receives the sub-carriers that are then weighted by the matrix UH of receive antenna weighting factors. This is expressed by the following equation, where U is N×min (M,N) and V is M×min (M, N), as U and V are unitary matrices.
                                                        y              =                                                                    U                    H                                    ⁢                  HVx                                +                                                      U                    H                                    ⁢                  n                                                                                                        =                                                                                          U                      H                                        ⁡                                          (                                              UDV                        H                                            )                                                        ⁢                  Vx                                +                                                      U                    H                                    ⁢                  n                                                                                                        =                                                                    (                                                                  U                        H                                            ⁢                      U                                        )                                    ⁢                                      D                    ⁡                                          (                                                                        V                          H                                                ⁢                        V                                            )                                                        ⁢                  x                                +                                                      U                    H                                    ⁢                  n                                                                                                        =                              IDIx                +                                                      U                    H                                    ⁢                  n                                                                                                        y              =                              Dx                +                                                      U                    H                                    ⁢                  n                                                                                        (        7        )            
Here, receive signal y and transmit signal x have (min (M, N)×1) vectors, not determined by the number of transmit antennas and the number of receive antennas.
Because D is the diagonal matrix, the transmit signal sub-carriers can be received without crosstalk. Since the amplitude of each of the independent MIMO channels is proportional to the square root of the eigenvalue λ for the channel, the power of each MIMO channel is proportional to λ.
Since the noise component n is also an eigenvector normalized to a norm of 1 in the U column, UHn does not affect the noise power. The size of UHn is a set of (min (M, N)) vectors, which is the same as the size of y and x.
In the SVD-MIMO transmission, in this way, logically independent multiple MIMO channels free of crosstalk can be available simultaneously in the same frequency band. Thus, using the same frequency band, multiple data streams can be transmitted simultaneously by wireless communication, and enhanced transmission speed can be achieved.
The number of MIMO channels available in the SVD-MIMO communication system matches the number of transmit antennas M or the number of receive antennas N, which is smaller, min [M, n]. The matrix V of transmit antenna weighting factors consists of as many transmit vectors vi as the number of MIMO channels (V=[v1, v2, . . . , vmin[M, N]]). The elements of the transmit vectors vi are as many as the number of transmit antennas M.
Generally, in the closed-loop MIMO scheme typified by SVD-MIMO, the transmitter is capable of calculating optimum weight factors for its antennas, based on information for the propagation paths. Furthermore, it is known that, by selecting an optimal coding ratio and a modulation scheme to be applied to bit streams on transmit antenna chains, more ideal information transmission can be realized.
However, practical operation of a system of the closed-loop MIMO scheme encounters such a problem that, if the conditions of the channels vary to a great extent, as the transmitter and the receiver move, feedback from the receiver to the transmitter must occur more frequently. In the SVD-MIMO communication scheme, it is not easy to calculate the singular value decomposition in real time. In addition, it is necessary to perform a setup procedure for advanced notification of V or UH obtained by the SVD calculation to the other end of communication.
By way of example, for an Orthogonal Frequency Division Multiplexing (OFDM) communication system of IEEE 802.11a, namely, in a 5-GHz band, one LAN system to which the SVD-MIMO transmission is applied, let us consider how much will be information of the matrix V of transmit antenna factors. Given that three transmit antenna elements and three receive antenna elements are employed, the matrix V of transmit antenna factors is 3×3, having nine elements. If one element is assumed consisting of a real number and a complex number which are accurate to 10 bits and the matrices V for 52 carriers are required, 9,360 bits (=9 (elements of the matrix)×2 (the real part and imaginary part of a complex number)×10 (bits)×52 (OFDM sub-carriers) must be fed back from the receiver to the transmitter.
A point that must be considered when constructing an actual SVD-MIMO transmission/reception system is now discussed.
In the basic form of the SVD-MIMO transmission scheme, at the receiver, by the singular value decomposition for the acquired channel matrix H, a set of receive weight vectors UH and a set of transmit weight vectors V that are employed at the transmitter are obtained, and this set of the vectors V is fed back to the transmitter. At the transmitter, this set of the vectors V is used as the set of the weights for transmission.
However, in the event that the amount of transmit weight matrix V information to be fed back to the transmitter is so large and sparsified V information is transmitted back, the orthogonal state of the MIMO channels will be altered due to errors from true V information and crosstalk will occur.
In view hereof, after the receiver feedbacks the transmit weight matrix V to the transmitter, usually, the transmitter transmits a reference signal weighted with the matrix V to the receiver and the receiver acquires the channel matrix again. Given that the channel matrix is H, the receiver can acquire a channel matrix HV from the reference signal weighted by V.
At the receiver, an inverse matrix of the HV is obtained and used as a set of weights for reception. Since H=UDVH, HV and its inverse will be obtained, as expressed in the equation below:
                                                        HV              =                                                UDV                  H                                ⁢                V                                                                                        =              UD                                                                                                            (                  HV                  )                                -                            =                                                (                  UD                  )                                -                                                                                        =                                                D                  -                                ⁢                                  U                  -                                                                                                        =                                                D                  -                                ⁢                                  U                  H                                                                                        (        8        )            
This is such that, after received sub-carriers are weighted with UH in the same manner as in normal SVD-MIMO, the separated streams for the MIMO channels are merely multiplied, respectively, by constants that are derived from the diagonal elements λi of the diagonal matrix D.
Arrangement in which the matrix V is used as a set of weights for transmission at the transmitter and the inverse matrix of HV is used as a set of weights for reception at the receiver is the same as performance of normal SVD-MIMO and V mismatch at the transmitter and the receiver does not occur. Therefore, this arrangement can be used practically.
[Patent document 1] JP-A No. H10-84324
[Non-patent document 1] International Standard ISO/IEC 8802-11:1999 (E) ANSI/IEEE Std 802.11, 1999 Edition, Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications
[Non-patent document 2] ETSI TS 101 761-1 V.1.2.1(2000-11) Broadband Radio Access Networks (BRAN); HIPERLAN Type 2; Data Link Control (DLC) Layer; Part 1: Basic Data Transport Functions
[Non-patent document 3] ETSI TS 101 761-2 V1.1.1 (2000-04) Broadband Radio Access Networks (BRAN); HIPERLAN Type 2; Data Link Control (DLC) Layer; Part 2: Radio Link Control (RLC) sublayer
[Non-patent document 4] Supplement to IEEE Standard for Information technology—Telecommunications and information exchange between systems—Local and metropolitan area networks—Specific requirements—Part 11: Wireless LN Medium Access Control (MAC) and Physical Layer (PHY) specifications: High-speed Physical Layer in the 5 GHz Band
[Non-patent document 5] http://web.archive.org/web/20031024-re/http://radio3.ee.uec.ac.jp/MIMO-IEICE-TB.pdf (as of Oct. 24, 2003)
To carry out SVD-MIMO communication, it is necessary to acquire the channel matrix or the like. Meantime, in typical wireless communication systems, the CSMA/CA scheme is applied for collision avoidance and, at the same time, the so-called RTS/CTS procedure is performed to get a transmission right for the purpose of, for instance, solving a hiding terminal problem. Therefore, channel matrix acquisition can be performed by using packets such as RTS, CTS, DATA, and ACK and through a control procedure that will be described below (see FIG. 6). It is assumed that the transmitter acquires a set of transmit weights V before the start of the RTS/CTS procedure.
(Step 1)
The transmitter transmits an RTS packet to the receiver. A reference signal is attached to the RTS packet.
(Step 2)
At the receiver, a channel matrix H is acquired from the RTS packet received.
(Step 3)
At the receiver, from the acquired channel matrix H, it is determined what modulation scheme is employed and how many independent spatial channels are available.
Upon receiving the RTS, the receiver may have requirement to determine a modulation scheme to be employed. For instance, the receiver may want to stop transmission from nodes or terminals in its vicinity until the completion of ACK, using a Network Allocation Vector (NAV) which should be attached to CTS. For Short NAV setting, it is necessary to perceive the modulation scheme employed for the channel and the bit rate on the channel to calculate time for transmitting and receiving data. To determine what modulation scheme should be used at the transmitter for transmitting data, it is necessary to know the conditions of the MIMO channels, namely, eigenvalues λ per channel, derived by singular value decomposition of the matrix H, so that what conditions of the MIMO channels will be perceived at the receiver.
(Step 4)
The receiver returns a CTS to the transmitter. To the CTS, a reference signal for channel matrix estimation is attached.
(Step 5)
At the transmitter, from the reference signal attached to the CTS transmitted from the receiver, a channel matrix H in the reverse direction is acquired.
If calibration is performed to compensate for difference in the characteristics of the analog circuits in the antenna chains of the transmitter and difference in the characteristics of the analog circuits in the antenna chains of the receiver, the transfer functions in the forward and reverse directions will be the same. A method for calibrating the difference in the characteristics of the analog circuit portions at the transmitter and the receiver is described, e.g., in JP-B (Japanese Examined Patent Application Publication) 2003-426294 whose patent right has already been assigned to the present inventors.
(Step 6)
The transmitter executes singular value decomposition of the acquired matrix H in the reverse direction and determines the weights V for transmission in the forward direction. Of course, the weights V for transmission in the forward direction, obtained by singular value decomposition at the receiver, may be fed back to the transmitter; however, its information amount is excessively large. Therefore, the receiver transmits back the reference signal having a small quantity of data and the transmitter acquires V as above.
(Step 7)
In response to the reception of the CTS signal from the receiver, the transmitter transmits a data packet. To the beginning of this data packet, a reference signal weighted by V is attached, followed by user data (payload).
(Step 8)
At the receiver, a channel matrix HV is acquired from the reference signal weighted by V, its inverse matrix (see equation (8)) is obtained as a set of weights for reception, and thereby weighted user data is received.
When the SVD-MIMO communication is performed, according to the above communication procedure, each time a data transmission process by the RTS/CTS procedure is performed, two times altogether of singular value decomposition calculation at the steps 3 and 6 and one time of inverse matrix calculation at the step 8 must be executed. This poses a problem of heavy calculation loads imposed on the transmitter and receiver. Especially, the singular value decomposition calculation consumes large power when it is executed, because it uses a considerable number of multipliers. Consequently, there is a need to decrease the quantity of singular value decomposition calculation, even just one time.